Bug#727708: Value of reading other's position statements
Miroslaw Baran <miroslaw+c+debbugs@makabra.org> writes:
> You wrote:
>> One non-feature of upstart which I happen to care strongly about is its
>> use of ptrace(2) to figure out what a job is doing. This destroys any
>> attempt to just use "strace foo" as the job, if one really needs to
>> figure out what a piece of software is doing wrong. Thanks but no
>> thanks.
> Let me allow to quote the upstart's position statement:
That statement talks about attaching gdb later in the lifetime of the
process. It doesn't address Matthias's point about running the daemon
itself under strace.
The issues discussed with valgrind here:
>> For valgrind, it's true that this would be less straightforward to do
>> as part of an upstart job. If there were widespread demand for solving
>> this, it wouldn't be difficult; but this doesn't seem like something
>> that has much practical impact. It's unlikely that someone using
>> valgrind for debugging will need to debug via an upstart job or systemd
>> unit, as opposed to running the service directly.
are, I believe, the same as the issues with strace, so I think this
position statement concedes that Matthias is correct and the use of ptrace
does prevent that particular use case.
I don't personally consider this a major issue, but it is a minor one. I
have personally done exactly that (replaced the daemon invocation with one
run under strace -o /root/tmp/trace) to debug problems in the past. You
can attach strace later, just like you can attach gdb later, but that
doesn't help if the problem you're trying to get a system call trace for
is during the daemon startup.
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: